2 * Copyright (c) 2000-2003 Tor Egge
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * $FreeBSD: src/sys/ufs/ffs/ffs_rawread.c,v 1.3.2.2 2003/05/29 06:15:35 alc Exp $
27 * $DragonFly: src/sys/vfs/ufs/ffs_rawread.c,v 1.28 2008/06/19 23:27:39 dillon Exp $
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/fcntl.h>
36 #include <sys/mount.h>
37 #include <sys/namei.h>
38 #include <sys/vnode.h>
40 #include <sys/filio.h>
41 #include <sys/ttycom.h>
47 #include <machine/limits.h>
49 #include <vm/vm_extern.h>
50 #include <vm/vm_object.h>
51 #include <sys/kernel.h>
52 #include <sys/sysctl.h>
54 static int ffs_rawread_readahead(struct vnode *vp, caddr_t udata, off_t offset,
55 size_t len, struct buf *bp);
56 static int ffs_rawread_main(struct vnode *vp,
59 static int ffs_rawread_sync(struct vnode *vp);
61 int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone);
63 void ffs_rawread_setup(void);
65 SYSCTL_DECL(_vfs_ffs);
67 static int ffsrawbufcnt = 4;
68 SYSCTL_INT(_vfs_ffs, OID_AUTO, ffsrawbufcnt, CTLFLAG_RD, &ffsrawbufcnt, 0,
69 "Buffers available for raw reads");
71 static int allowrawread = 1;
72 SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0,
73 "Flag to enable raw reads");
75 static int rawreadahead = 1;
76 SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0,
77 "Flag to enable readahead for long raw reads");
81 ffs_rawread_setup(void)
83 ffsrawbufcnt = (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8;
88 ffs_rawread_sync(struct vnode *vp)
93 * Check for dirty mmap, pending writes and dirty buffers
95 lwkt_gettoken(&vp->v_token);
96 if (bio_track_active(&vp->v_track_write) ||
97 !RB_EMPTY(&vp->v_rbdirty_tree) ||
98 (vp->v_flag & VOBJDIRTY) != 0) {
99 /* Attempt to msync mmap() regions to clean dirty mmap */
100 if ((vp->v_flag & VOBJDIRTY) != 0) {
101 struct vm_object *obj;
102 if ((obj = vp->v_object) != NULL)
103 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC);
106 /* Wait for pending writes to complete */
107 error = bio_track_wait(&vp->v_track_write, 0, 0);
111 /* Flush dirty buffers */
112 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
113 if ((error = VOP_FSYNC(vp, MNT_WAIT, 0)) != 0) {
116 if (bio_track_active(&vp->v_track_write) ||
117 !RB_EMPTY(&vp->v_rbdirty_tree))
118 panic("ffs_rawread_sync: dirty bufs");
124 lwkt_reltoken(&vp->v_token);
130 ffs_rawread_readahead(struct vnode *vp, caddr_t udata, off_t loffset,
131 size_t len, struct buf *bp)
140 bsize = vp->v_mount->mnt_stat.f_iosize;
143 * Make sure it fits into the pbuf
145 iolen = (int)(intptr_t)udata & PAGE_MASK;
146 if (len + iolen > bp->b_kvasize) {
153 * Raw disk address is in bio2, but we wait for it to
156 bp->b_flags &= ~B_ERROR;
157 bp->b_loffset = loffset;
158 bp->b_bio2.bio_offset = NOOFFSET;
159 bp->b_bio1.bio_done = biodone_sync;
160 bp->b_bio1.bio_flags |= BIO_SYNC;
162 blockoff = (loffset % bsize) / DEV_BSIZE;
164 error = VOP_BMAP(vp, bp->b_loffset, &bp->b_bio2.bio_offset,
165 &bforwards, NULL, BUF_CMD_READ);
168 dp = VTOI(vp)->i_devvp;
169 if (bp->b_bio2.bio_offset == NOOFFSET) {
171 * Fill holes with NULs to preserve semantics
173 if (len + blockoff * DEV_BSIZE > bsize)
174 len = bsize - blockoff * DEV_BSIZE;
176 if (vmapbuf(bp, udata, len) < 0)
180 bzero(bp->b_data, bp->b_bcount);
182 /* Mark operation completed (similar to bufdone()) */
188 if (len + blockoff * DEV_BSIZE > bforwards)
189 len = bforwards - blockoff * DEV_BSIZE;
190 bp->b_bio2.bio_offset += blockoff * DEV_BSIZE;
192 if (vmapbuf(bp, udata, len) < 0)
196 * Access the block device layer using the device vnode (dp) and
197 * the translated block number (bio2) instead of the logical block
200 * Even though we are bypassing the vnode layer, we still
201 * want the vnode state to indicate that an I/O on its behalf
204 bp->b_cmd = BUF_CMD_READ;
205 bio_start_transaction(&bp->b_bio1, &vp->v_track_read);
206 vn_strategy(dp, &bp->b_bio2);
211 ffs_rawread_main(struct vnode *vp, struct uio *uio)
214 struct buf *bp, *nbp, *tbp;
220 udata = uio->uio_iov->iov_base;
221 resid = uio->uio_resid;
222 offset = uio->uio_offset;
232 if (bp == NULL) { /* Setup first read */
233 /* XXX: Leave some bufs for swap */
234 bp = getpbuf_kva(&ffsrawbufcnt);
235 error = ffs_rawread_readahead(vp, udata, offset,
240 if (resid > bp->b_bufsize) { /* Setup fist readahead */
241 /* XXX: Leave bufs for swap */
242 if (rawreadahead != 0)
243 nbp = trypbuf_kva(&ffsrawbufcnt);
247 nerror = ffs_rawread_readahead(
249 udata + bp->b_bufsize,
250 offset + bp->b_bufsize,
251 resid - bp->b_bufsize,
254 relpbuf(nbp, &ffsrawbufcnt);
261 biowait(&bp->b_bio1, "rawrd");
265 iolen = bp->b_bcount - bp->b_resid;
266 if (iolen == 0 && (bp->b_flags & B_ERROR) == 0) {
267 nerror = 0; /* Ignore possible beyond EOF error */
271 if ((bp->b_flags & B_ERROR) != 0) {
275 clearbiocache(&bp->b_bio2);
279 if (iolen < bp->b_bufsize) {
280 /* Incomplete read. Try to read remaining part */
281 error = ffs_rawread_readahead(
283 bp->b_bufsize - iolen, bp);
286 } else if (nbp != NULL) { /* Complete read with readahead */
292 clearbiocache(&nbp->b_bio2);
294 if (resid <= bp->b_bufsize) { /* No more readaheads */
295 relpbuf(nbp, &ffsrawbufcnt);
297 } else { /* Setup next readahead */
298 nerror = ffs_rawread_readahead(
299 vp, udata + bp->b_bufsize,
300 offset + bp->b_bufsize,
301 resid - bp->b_bufsize,
304 relpbuf(nbp, &ffsrawbufcnt);
308 } else if (nerror != 0) {/* Deferred Readahead error */
310 } else if (resid > 0) { /* More to read, no readahead */
311 error = ffs_rawread_readahead(vp, udata, offset,
319 relpbuf(bp, &ffsrawbufcnt);
320 if (nbp != NULL) { /* Run down readahead buffer */
321 biowait(&nbp->b_bio1, "rawrd");
323 relpbuf(nbp, &ffsrawbufcnt);
328 uio->uio_iov->iov_base = udata;
329 uio->uio_resid = resid;
330 uio->uio_offset = offset;
336 ffs_rawread(struct vnode *vp,
340 if (allowrawread != 0 &&
341 uio->uio_iovcnt == 1 &&
342 uio->uio_segflg == UIO_USERSPACE &&
343 uio->uio_resid == uio->uio_iov->iov_len &&
344 (curthread->td_flags & TDF_DEADLKTREAT) == 0) {
345 int secsize; /* Media sector size */
346 off_t filebytes; /* Bytes left of file */
347 int blockbytes; /* Bytes left of file in full blocks */
348 int partialbytes; /* Bytes in last partial block */
349 int skipbytes; /* Bytes not to read in ffs_rawread */
354 /* Only handle sector aligned reads */
356 secsize = ip->i_devvp->v_rdev->si_bsize_phys;
357 if ((uio->uio_offset & (secsize - 1)) == 0 &&
358 (uio->uio_resid & (secsize - 1)) == 0) {
360 /* Sync dirty pages and buffers if needed */
361 error = ffs_rawread_sync(vp);
365 /* Check for end of file */
366 if (ip->i_size > uio->uio_offset) {
367 filebytes = ip->i_size - uio->uio_offset;
369 /* No special eof handling needed ? */
370 if (uio->uio_resid <= filebytes) {
372 return ffs_rawread_main(vp, uio);
375 partialbytes = ((unsigned int) ip->i_size) %
377 blockbytes = (int) filebytes - partialbytes;
378 if (blockbytes > 0) {
379 skipbytes = uio->uio_resid -
381 uio->uio_resid = blockbytes;
382 error = ffs_rawread_main(vp, uio);
383 uio->uio_resid += skipbytes;
386 /* Read remaining part using buffer */